Anthropic needs to bring in Amazon-style earnings to justify its $2 trillion valuation—but it’s barely turned a profit

Anthropic backers expect an IPO in October with a targeted valuation of $2T+, eclipsing SpaceX’s $1.77T and more than doubling the $965B value from its May Series H. The push faces a profitability hurdle: peers’ typical ~34x trailing and 25x forward earnings multiples imply $59B–$79B of annual net profits would be needed, yet the WSJ reports Q2’26 revenue rising to $10.9B with an operating profit (not net income). Separately, Bloomberg says Anthropic is in talks to buy Decart AI for $6B, which could help bolster the business pre-debut, while the valuation dynamic may pressure OpenAI’s post-2027 IPO framing.

Analysis

The real market event is not an IPO headline; it is the creation of a live, public comp for frontier-model economics. That tends to help listed holders with embedded marks and strategic ties first, but it also forces the whole AI complex to answer a harder question: which revenues are recurring software economics versus subsidized compute consumption. In that sense, AMZN has the cleanest near-term optical upside, while GOOGL and MSFT face a more subtle pressure to show that their own AI businesses can scale without permanent margin dilution.

Second-order winners are the picks-and-shovels names that monetize model scale without needing the market to believe in terminal software margins. AVGO and TSM are better positioned than NVDA if the next leg of spend shifts toward custom silicon, networking, and inference optimization rather than raw GPU brute force. The Decart angle matters because efficiency improvements can lengthen useful chip life and lower marginal compute intensity; that is bullish for adoption but not necessarily for the highest-multiple compute supplier.

The main risk is that public investors anchor on a $2T valuation and then immediately demand profits, not just growth, which could turn the IPO into a sentiment peak for AI beta rather than a durable rerating. Over the next 1-3 months, the S-1 will matter far more than the price talk; the key falsifier for the bullish thesis is disclosure showing short-dated compute commitments, heavy customer concentration, or weak net income conversion. Over 6-18 months, the structural issue is whether frontier labs become capital-efficient software franchises or perpetual infrastructure renters.

Key Decisions for Investors

• Long AMZN / short NVDA as a 1-3 month relative-value trade into the filing window; thesis is that AMZN gets the cleaner mark-to-market and strategic cloud benefit, while NVDA is most exposed if the market starts pricing AI efficiency over raw token demand. Risk/reward: attractive only if the S-1 emphasizes optimization and enterprise economics; cut if GPU demand reaccelerates or hyperscaler capex guidance surprises higher.
• Build a modest long AVGO / TSM basket on pullbacks over the next 1-3 months; these are the better second-order beneficiaries if frontier-model spend shifts toward custom silicon, interconnect, and foundry throughput. Falsifier: disclosure that Anthropic remains overwhelmingly spot-GPU dependent with no meaningful custom hardware mix.
• Use any IPO-related pop in CRM and GOOGL to trim exposure rather than add; the market may temporarily overpay for 'enterprise AI validation,' but the cash-flow translation is likely 12-18 months away. This is especially true if the filing shows customer concentration or service-heavy revenue that is expensive to support.
• If the S-1 confirms strong net-margin conversion and durable enterprise retention, rotate from broad AI beta into AMZN and AVGO instead of chasing the headline. The upside case is real, but the cleanest public-market expression is ownership of the infrastructure monetizers, not the private-lab valuation itself.